Evaluation of geographic variations in cancer rates may suggest clues to the roles of environmental or cultural influences. The identification of regions at notably high or low risk may indicate areas where more intensive studies might be particularly fruitful. We have completed preparations for the production of a new atlas of cancer mortality in the United States during 1970-94. We edited data files pertaining to more than 9.5 million whites and 1.1 million blacks who died from cancer and prepared the corresponding required estimates of person-years at risk. We developed software referred to as the "Mortality Rate Generator" program to calculate observed and expected counts, age-adjusted rates, and confidence limits by cancer, race, sex, time period, and geographic area. Using special graphics software for Windows on a personal computer, we have generated more than 140 maps at the county or State Economic Area level that will be included in the atlas. Summary tables and figures have been prepared. The analysis of these data and maps is in progress and the accompanying text is in the final stages of development. We are investigating methods to make the rates and maps available on the internet, such as through the Division of Cancer Epidemiology and Genetics website http://www-dceg.ims.nci.nih.gov/MRG/index.html. We are starting to analyze changes in the geographic patterns for those cancers with substantial variation by area, race, and/or gender.